Risk of tuberculosis is higher with anti–tumor necrosis factor monoclonal antibody therapy than with soluble tumor necrosis factor receptor therapy: The three-year prospective french research axed on tolerance of biotherapies registry†
Université Paris 7 Denis Diderot; INSERM U738; Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Bichat, Biostatistique et Recherche Clinique, Paris, France
Tuberculosis (TB) is associated with anti–tumor necrosis factor (anti-TNF) monoclonal antibody (mAb) therapy, but whether this association is drug-specific remains a concern. Our objective was to describe cases of TB associated with anti-TNF mAb therapy, identify risk factors, and estimate the incidence.
We conducted an incidence study and a case–control analysis to investigate the risk of newly diagnosed TB associated with the use of anti-TNF agents. As part of the French Research Axed on Tolerance of Biotherapies (RATIO) registry, for 3 years we collected cases of TB among French patients receiving anti-TNF mAb therapy for any indication; for each case, 2 patients treated with anti-TNF agents served as control subjects.
We collected 69 cases of TB in patients treated for rheumatoid arthritis (n = 40), spondylarthritides (n = 18), inflammatory colitis (n = 9), psoriasis (n = 1) and Behçet's disease (n = 1) with infliximab (n = 36), adalimumab (n = 28), and etanercept (n = 5). None of the patients had received correct chemoprophylactic treatment. The sex- and age-adjusted incidence rate of TB was 116.7 per 100,000 patient-years. The standardized incidence ratio (SIR) was 12.2 (95% confidence interval [95% CI] 9.7–15.5) and was higher for therapy with infliximab and adalimumab than for therapy with etanercept (SIR 18.6 [95% CI 13.4–25.8] and SIR 29.3 [95% CI 20.3–42.4] versus SIR 1.8 [95% CI 0.7–4.3], respectively). In the case–control analysis, exposure to infliximab or adalimumab versus etanercept was an independent risk factor for TB (odds ratio [OR] 13.3 [95% CI 2.6–69.0] and OR 17.1 [95% CI 3.6–80.6], respectively). Other risk factors were age, the first year of anti-TNF mAb treatment, and being born in an endemic area.
The risk of TB is higher for patients receiving anti-TNF mAb therapy than for those receiving soluble TNF receptor therapy. The increased risk with early anti-TNF treatment and the absence of correct chemoprophylactic treatment favor the reactivation of latent TB.
Treatment with tumor necrosis factor (TNF) antagonists has been recognized as a risk factor for active tuberculosis (TB) in patients with immune-mediated inflammatory diseases, including rheumatoid arthritis (RA), ankylosing spondylitis (AS), Crohn's disease, psoriatic arthritis, and psoriasis (1–7). Most cases of TB develop soon after treatment initiation and correspond to reactivation of a latent TB infection (4, 5). Scientific organizations and health authorities worldwide have proposed recommendations for screening patients with latent TB infection and treating such patients before initiating anti-TNF treatment, and the effectiveness of these recommendations has been demonstrated (8).
Treatment with all 3 available TNF antagonists has been associated with an increased incidence of TB. However, a possible difference between infliximab and etanercept in the incidence of TB is suspected from the spontaneous reporting of TB cases to the US Food and Drug Administration (FDA) (5) before September 2002, when no recommendations for screening for latent TB existed. In one study, the incidence of TB was 28 per 100,000 patient-years for patients receiving etanercept versus 54 per 100,000 patient-years for patients receiving infliximab (9). Likewise, 4 registries showed small differences (that were not statistically significant) in the incidence of TB between patients with RA receiving infliximab and those receiving etanercept; perhaps because in each of these registries the number of TB cases was low (n = 2–13), these registries were not powered enough to demonstrate very rare events or to determine a difference in risk depending on the anti-TNF agent (1, 2, 6, 7). No specific epidemiologic data are available for adalimumab, but results of phase III and phase IV clinical studies suggested that treatment with this drug may increase the risk of TB to a level close to that associated with infliximab treatment (10).
Thus, we used another strategy to investigate the incidence of TB by setting up a national registry to collect all cases of TB occurring during 3 years in French patients receiving anti-TNF monoclonal antibody (mAb) therapy, regardless of the indication for use. The aim of this large prospective study was to describe cases of TB associated with anti-TNF mAb therapy and their outcome, to identify risk factors for TB occurrence in patients receiving anti-TNF mAb therapy, and to estimate the global and drug-specific incidence of TB after the establishment and widespread knowledge by clinicians of recommendations for TB prophylaxis before the initiation of anti-TNF mAb therapy.
PATIENTS AND METHODS
The French Research Axed on Tolerance of Biotherapies (RATIO) registry was designed by a multidisciplinary group to prospectively collect all cases of TB occurring from February 1, 2004 to January 1, 2007 in patients who were receiving anti-TNF mAb therapy (11). This registry involved clinicians from all concerned medical specialties, the French regulatory agency for medical products (Agence Française de Sécurité Sanitaire des Produits de Santé [AFSSAPS]), and its network of 31 regional pharmacovigilance centers. Clinicians who reported at least 1 case associated with anti-TNF mAb therapy to the RATIO registry are shown in Appendix A.
We conducted an incidence study and a case–control analysis to investigate the risk of newly diagnosed TB associated with treatment using anti-TNF agents.
French guidelines (12) elicited in 2002 recommended anti-TB chemoprophylaxis before all anti-TNF mAb therapy for patients presenting with at least 1 of the following: diameter of tuberculin skin test reaction 10 mm (revised to 5 mm in July 2005), abnormal chest radiography results with calcifications >1 cm, previous untreated exposure to TB, or episode of TB. Correct chemoprophylaxis was defined as a 9-month course of treatment with isoniazide or a 3-month course of treatment with 2 anti-TB drugs, including rifampin (12).
Identification and validation of TB cases.
To enhance the exhaustiveness of the collection of cases of TB, different sources were used. All cases reported to the 31 French regional pharmacovigilance centers were collected (reporting of severe drug-related adverse events is mandatory for clinicians in France), along with cases reported directly to the companies marketing anti-TNF agents.
In addition, physicians from all French hospital centers who were involved in either the prescription of TNF blockers (i.e., those in the rheumatology, internal medicine, gastroenterology, and dermatology departments) or the management of TB (i.e., those in infectious diseases centers, intensive care units, and chest medicine units) were directly required to report each newly diagnosed case. A direct mail reminder sent 4 times yearly and several communications at congresses or in specialized press encouraged these physicians to report cases.
We also obtained cases from the Institut National de Veille Sanitaire, the French national public health surveillance system that monitors TB in France with a mandatory clinician- and biologist-based notification system.
A detailed standardized case report form was completed for each reported case of TB (whatever the source of notification), and additional documents were collected (hospitalization summary, results of histologic and microbiologic assessments, and data, if necessary) for use in the validation process.
Validation of cases.
Included in the RATIO registry were all cases (from all sources) with a validated diagnosis of TB according to the International Classification of Diseases, Tenth Revision (A15–A19). An expert committee comprising 3 experts on TB (DS, SB, and OL) validated cases by consensus on the basis of the detailed standardized case report form and the additional documents.
Assessment of exposure to anti-TNF mAb therapy.
Infliximab has been available since 1999, etanercept has been available since 2002, and adalimumab has been available since 2004. In France, it is mandatory that the first prescription for any of these agents is from a hospital-based physician. For patients who received >1 anti-TNF agent, we considered 3 ways of addressing the exposure, as follows: first anti-TNF agent, last anti-TNF agent, and any use of the anti-TNF agent.
Incidence rate of TB.
We estimated the annual incidence rate of TB in patients treated with anti-TNF mAb therapy, adjusted for age and sex, with the French population used as the reference. The numerator consisted of the validated cases of TB from the RATIO registry. For the denominator, we estimated the number of patient-years of receipt of anti-TNF agents in France during the 3-year period of the study (2004, 2005, and 2006), as provided by the different sources. For this purpose, we obtained the total number of doses of each anti-TNF agent sold in France during the 3-year study period, according to both the AFSSAPS and the 3 pharmaceutical firms (Abbott, Schering-Plough, and Wyeth). Each firm provided an estimation of the number of patient-years for the 3 anti-TNF agents on the basis of the number of doses sold and the mean dosage used for each indication. These data were very consistent. We used the mean of these different estimates to derive an estimation of the number of patients receiving each anti-TNF agent, which we used in the main analysis.
In addition, for the sensitivity analysis, we obtained independent data from the Régime Social des Indépendants (RSI), the French health insurance fund for self-employed workers, which includes 3 million enrollees (i.e., 5% of the French population representative of the whole population) and provides claims and demographic data for the use of etanercept and adalimumab.
Standardized incidence ratios (SIRs).
To assess the distribution of underlying diseases as well as the sex and age of treated patients, we used surveys from prescribing clinicians, hospital pharmacies, and nonhospital pharmacies performed by the firms and the RSI files for the patients receiving etanercept and adalimumab.
On the basis of the previous data and the French incidence rate of TB from 2002 to 2006, by both 10-year age and sex classes (from the Institut National de Veille Sanitaire) (13), we estimated the SIR (14) to compare the TB risk associated with anti-TNF treatment with that in the French population.
Risk of TB in patients receiving anti-TNF mAb therapy.
A case–control study was performed, using cases from the RATIO registry.
All cases were validated cases of TB in the registry with a labeling indication for the use of anti-TNF treatment (i.e., RA, spondylarthropathy [SpA; AS or psoriatic arthritis], ulcerative colitis or Crohn's disease, or psoriasis).
TB-free patients receiving anti-TNF treatment for a labeling indication were included from centers involved in the RATIO registry in a global pool of control subjects. From that pool, we randomly selected patients for a database of control subjects reflecting the proportion of patients in France receiving each of the 3 anti-TNF drugs. Two controls per case were randomly matched by sex and underlying inflammatory disease from this database of control subjects. We also used a second sample of control subjects who were randomly selected from the same database of controls, with the same matching criteria (second matching). In addition, we used a third sample of control subjects (1 control per case), adding the time from onset of anti-TNF treatment as matching criteria (third matching).
The number of cases of TB in France during the study period determined the sample size. A descriptive analysis was performed for the whole sample.
Annual incidence rate of TB.
The SIR was calculated for anti-TNF agent use as a whole and for agents used individually. The 95% confidence intervals (95% CIs) for each SIR were calculated, and the deviation of the SIR from 1.00 was tested, assuming a Poisson distribution for the observed cases because of the small number of cases.
We identified the risk factors for TB by both univariate and multivariate analysis (conditional logistic regression model). The predefined factors potentially predictive of the occurrence of TB considered were age, being born in a TB-endemic area (according to the 2008 World Health Organization [WHO] Report WHO/HTM/TB/2008.393), duration of the underlying inflammatory disease, diameter of the tuberculin skin test reaction at screening, chest radiography results at screening indicating TB sequelae, last anti-TNF agent received, time from the onset of anti-TNF treatment, use of a potentially immunosuppressive disease-modifying antirheumatic drug during the last year, use of steroids during the last year, and use of >10 mg/day or boluses of steroids during the last year.
Factors included in the stepwise multivariate model were those associated with the status (case or control) in univariate analysis with a significance level of P < 0.20. We used the bootstrap method to examine the stability of our final model (1,000 replicated bootstrap samples to identify candidate variables that were more often retained). P values less than 0.05 were considered significant.
We performed subgroup analyses for the more frequent underlying inflammatory diseases, RA and SpA.
To better assess the drug-specific risk, we performed sensitivity analyses for the SIR and the odds ratio (OR) measuring the risk associated with adalimumab or infliximab versus etanercept, varying the definition of exposure (all anti-TNF agents, by anti-TNF agent, and for patients receiving only 1 anti-TNF agent, patients receiving anti-TNF mAb therapy for at least 6 weeks, and patients currently receiving anti-TNF mAb therapy), the different denominator estimates (denominator estimates from each of the 3 pharmaceutical firms and from the RSI), and the different criteria used for matching the controls with cases (second and third matching).
The main analyses were independently performed by 2 statisticians. Statistical analysis was performed using SAS version 9.1 software (SAS Institute, Cary, NC).
Compliance with research ethics standards.
This study was authorized by the ethics committee of AP-HP, GHU Nord (institutional review board of Paris North Hospitals, Paris 7 University, AP-HP; authorization no. 162-08).
Description of the TB cases.
We collected data on 75 cases of TB, and 69 cases were validated. The characteristics of the 69 patients with a validated diagnosis of TB are shown in Table 1. In addition, 1 patient had diabetes, 3 had a history of cancer, none was human immunodeficiency virus positive, and 1 had a chronic obstructive pulmonary disease. Among the 58 patients receiving only 1 anti-TNF agent, 34 (58.6%) had received infliximab, 23 (39.7%) had received adalimumab, and 1 (1.7%) had received etanercept. Two patients (and 4 control subjects) had also received anakinra.
Table 1. Characteristics of the 69 patients with tuberculosis*
Except where indicated otherwise, values are the number (%). DMARDs = disease-modifying antirheumatic drugs.
Time from the initiation of the first/last anti–tumor necrosis factor (anti-TNF) treatment to the first symptoms of tuberculosis.
At least 1 of the following: mycophenolate mofetil, chlorambucil, cyclosporine, cyclophosphamide, mercaptopurine, methotrexate, leflunomide, azathioprine, abatacept, anakinra, certolizumab, CTLA-4Ig, rituximab.
Age, mean ± SD (median) years
58.5 ± 15.4 (61.0)
Born in a tuberculosis-endemic area
Underlying inflammatory disease
Crohn's disease or ulcerative colitis
Duration of underlying inflammatory disease, mean ± SD (median) years
11.2 ± 8.8 (8.3)
Only 1 anti-TNF agent received
Two anti-TNF agents received successively
Three anti-TNF agents received successively
Last anti-TNF agent received
Ever used adalimumab
Ever used etanercept
Ever used infliximab
Time since first anti-TNF treatment start, mean ± SD (median) months†
15.9 ± 14.4 (12.0)
Time since last anti-TNF treatment start, mean ± SD (median) months†
Risk factors for TB before anti-TNF mAb therapy were available for 51 patients. Thirty-four patients (66.7%) had at least 1 risk factor (diameter of tuberculin skin test reaction ≥5 mm, chest radiography results showing signs of a history of TB, history of untreated TB or exposure to TB, or being born in a TB-endemic area). The diameter of the tuberculin skin test reaction, which was available in 45 patients, was >10 mm in 4 patients (9%), 5–10 mm in 11 patients (24%), and <5 mm in 30 patients (67%). For 6 (10.5%) of 57 patients, the chest radiography findings suggested a history of TB (chest radiography was not performed in 12 patients). In addition, 10 patients had a history of exposure to TB, 1 patient a personal history of TB, 3 patients had a history of primary TB infection, and 14 patients were born in a TB-endemic area.
According to French guidelines for TB chemoprophylaxis before initiation of anti-TNF mAb therapy, with a tuberculin skin test reaction diameter cutoff of 5 mm or 10 mm, 26 of 46 (56.5%) or 17 of 43 (39.5%) of the patients, respectively, should have received chemoprophylaxis. None of the patients had received correct chemoprophylaxis according to the French recommendations (12) (3 had received incomplete chemoprophylaxis: 2 with a tuberculin skin test reaction diameter >10 mm and 1 without tuberculin skin testing before anti-TNF treatment), and TB developed 15–34 months later.
Clinical features were restricted to pleuropulmonary disease and/or mediastinal adenitis in 27 patients (39%) and were extrapulmonary in 42 patients (61%), as follows: disseminated disease in 28 patients (41%), nonmediastinal adenitis in 7 patients (10%), and other diseases (1 spondylodiscitis, 2 meningoencephalitis, 1 subcutaneous TB, 1 laryngeal TB, and 2 ileocolic TB) in 7 patients (10%). The diagnosis of TB was based on microbiologic evidence (direct Ziehl staining showing acid-fast bacilli and/or Mycobacterium tuberculosis isolation) in 47 patients, histologic evidence in 14 patients, clinical suspicion and positive QuantiFERON (Cellestis, Carnegie, Victoria, Australia) test results in 1 patient, and clinical suspicion associated with response to anti-TB treatment in 7 patients.
Three patients required hospitalization in an intensive care unit. After a median followup period of 22.9 months, 2 patients died (1 patient had disseminated TB and died before diagnosis, and the other patient died of a cause unrelated to TB), 45 were considered cured, 5 were still receiving treatment, and 5 experienced a relapse of TB (after a median delay of 9.4 months). The outcomes in 12 cases were missing. None of the 5 patients who experienced a relapse had restarted anti-TNF therapy before the relapse.
At the time of the TB diagnosis, anti-TNF mAb therapy was not stopped in 2 patients (both of whom were treated with infliximab). Moreover, anti-TNF mAb therapy was restarted in 8 patients (etanercept in 3, infliximab in 5) after a median anti-TB treatment duration of 5.8 months (range 5 days to 10 months). No recurrence of TB was observed in these 10 patients (median followup 17.7 months, range 8.2–32.2 months).
Time to occurrence of TB in patients receiving anti-TNF mAb therapy.
The median time to occurrence of TB since the start of anti-TNF treatment was 12.0 months. The median time since the start of the last anti-TNF agent to the occurrence of TB was 9.9 months. In 4 patients, TB had occurred despite discontinuation of anti-TNF mAb therapy 7.1–13.9 months previously. Among these 4 patients, the anti-TNF agent was infliximab in 3 and adalimumab in 1. As shown in Figure 1, the incidence of TB was higher during the first year of anti-TNF treatment than afterward and differed depending on the type of treatment: monoclonal antibodies (infliximab or adalimumab) or soluble TNF receptor (sTNFR; etanercept).
Incidence and risk of TB in patients receiving anti-TNF mAb therapy versus the general population.
The main analysis relied on a total of 57,711 patient-years of use of anti-TNF mAb therapy during the 2004–2006 period, with 18% of patients receiving adalimumab, 51% receiving etanercept, and 31% receiving infliximab serving as the denominator of the incidence rate. The annual incidence rate of TB, adjusted for age and sex, among patients receiving anti-TNF mAb therapy, with the French population as a reference, was 116.7 (95% CI 10.6–222.9 per 100,000 patient-years). The SIR was 12.2 (95% CI 9.7–15.5; P < 0.0001).
The risk of TB for patients receiving anti-TNF mAb therapy as compared with the French population differed depending on the anti-TNF agent used. The annual adjusted incidence rate of TB was 9.3 (range 0.0–9.4) per 100,000 for patients receiving etanercept, 187.5 (range 0.1–374.8) per 100,000 for patients receiving infliximab, and 215.0 (range 0.0–521.7) per 100,000 for patients receiving adalimumab, compared with 8.7 per 100,000 for the general French population (13). The SIR was 1.8 (95% CI 0.7–4.3; P = 0.20) for patients receiving etanercept, 18.6 (95% CI 13.4–25.8; P < 0.0001) for patients receiving infliximab, and 29.3 (95% CI 20.3–42.4; P < 0.0001) for patients receiving adalimumab. Because the incidence rates and SIRs for adalimumab and infliximab were close, and because their mechanisms of action are similar, we pooled data for adalimumab and infliximab for subgroup and sensitivity analyses. Still, a difference between anti-TNF mAb and sTNFR therapy was observed in the main analysis and in the different sensitivity analyses (Figure 2), even when we separately used the different estimates from independent sources, which gave very consistent adjusted incidence rates and SIRs (supplemental material is available from the corresponding author).
TB risk factors for patients receiving anti-TNF mAb therapy.
The case–control study involved 68 cases (as described in Patients and Methods, the patient treated for Behçet's disease was not included in this case–control study) and 136 control subjects (from a pool of control subjects whose anti-TNF treatments reflected those of all treated persons nationally, i.e., 18% receiving adalimumab, 51% receiving etanercept, and 31% receiving infliximab). The results of the univariate case–control analysis are shown in Table 2. In the multivariate analysis, the following factors were predictive of TB: age, being born in an endemic area, time from onset of anti-TNF treatment, and therapy with adalimumab or infliximab versus etanercept.
Table 2. TB risk factors in patients and control subjects receiving anti-TNF therapy, by univariate analysis*
TB patients (n = 68)
Controls (n = 136)
OR (95% CI)
Except where indicated otherwise, values are the number (%). 95% CI = 95% confidence interval; DMARD = disease-modifying antirheumatic drug.
Odds ratio (OR) for a 10-year increase.
For patients, time from the initiation of the last anti–tumor necrosis factor (anti-TNF) treatment until the first symptoms of tuberculosis (TB); for control subjects, time from the initiation of the last anti-TNF treatment until the last inclusion.
At least 1 of the following: mycophenolate mofetil, chlorambucil, cyclosporine, cyclophosphamide, mercaptopurine, methotrexate, leflunomide, azathioprine, abatacept, anakinra, certolizumab, CTLA-4Ig, rituximab.
Steroids >10 mg/day or boluses during the last year
The results of the main multivariate analysis are presented in Table 3. The risk of TB was higher with infliximab or adalimumab than with etanercept (OR 13.3, 95% CI 2.6–69.0 [P = 0.002]) and 17.1 (95% CI 3.6–80.6 [P = 0.0003]), respectively. As for the SIRs, we pooled data for adalimumab and infliximab for the subgroup and sensitivity analyses, for more power. The drug-specific ORs were also similar with the second and third matchings and in subgroup and sensitivity analyses (Figure 3).
Table 3. TB risk factors in 68 patients and 136 control subjects receiving anti-TNF therapy, by multivariate analysis*
OR (95% CI)
95% CI = 95% confidence interval.
Odds ratio (OR) for a 10-year increase.
For patients, time from the initiation of the last anti–tumor necrosis factor (anti-TNF) treatment until the first symptoms of tuberculosis (TB); for control subjects, time from the initiation of the last anti-TNF treatment until inclusion.
Our analysis of 69 cases of TB collected over 3 years in the RATIO registry clearly demonstrated a difference in the risk of TB between patients receiving anti-TNF mAb therapy and those receiving sTNFR therapy. We observed a 7–17-fold difference in risk depending on the 2 types of anti-TNF agents used, from 2 different methodologic approaches to assess the risk of TB: a comparison of the incidence rate with that in the French general population and a case–control study with controls receiving anti-TNF mAb therapy for the same underlying inflammatory diseases. The risk of TB was higher during the first year of anti-TNF treatment, which favored the reactivation of latent TB. Moreover, none of the 69 cases had been treated with correct chemoprophylaxis against TB before anti-TNF treatment was initiated.
After the FDA alert regarding the risk of TB associated with the use of infliximab, numerous countries set up registries to investigate the safety of anti-TNF agents. All registries of biologic agents except ours are cohort studies involving only a part of the population of focus and thus are not powered enough to demonstrate very rare events or investigate a difference in risk depending on the anti-TNF agent used, nor are the randomized controlled trials (15). In the US National Data Bank for Rheumatic Diseases cohort, the rate of TB was not increased in anti-TNF–naive patients with RA (6.2 [range 1.6–34.4] per 100,000 patient-years) but was increased in RA patients receiving infliximab (52.5 [range 14.3–134.4] per 100,000 patient-years) (7). However, in this study, the rate was calculated from only 4 cases of TB among 6,460 patients receiving infliximab. In the same study, no case of TB was observed among 2,327 patients receiving etanercept. In the British registry, the incidence of TB was 1.5 per 1,000 patient-years with infliximab and 0.5 per 1,000 patient-years with etanercept, but the numbers of cases were only 7 and 2, respectively (2). In the Swedish registry, the incidence rates were 1.5 per 1,000 patient-years with infliximab and 0.8 per 1,000 patient-years with etanercept; 9 cases were treated with infliximab alone, and 4 cases were treated with etanercept alone (1). Finally, among Korean patients, 2 cases of TB were observed among 90 patients receiving infliximab, and no case was observed among 103 patients receiving etanercept (6). The difference in risk between the types of anti-TNF agents was only suggested in these registries but was clearly demonstrated in the RATIO registry.
In our registry, patients were included when the adverse event occurred, not when the therapy began. Thus, our population of focus was the entire French population, and this allowed us to collect in 3 years many more cases of TB than were collected by all of the other registries. Furthermore, as far as we know, ours is the only biologic registry collecting safety data for patients receiving anti-TNF mAb therapy for any indication.
Contrary to the standard pharmacovigilance notifications, all of the files for TB patients included in the RATIO registry were reviewed and validated by an expert committee, which is a main quality issue, especially for the diagnosis of TB, which may be difficult.
As always, incidence rates might be somewhat underestimated or overestimated because of an imprecise numerator or denominator. Concerning the numerator, passive surveillance data may underestimate the incidence of adverse events, and thus the annual rate of TB in the RATIO registry might be toward the low end of the true incidence rate (16, 17). The denominator of the incidence rate was estimated only. The pharmaceutical firms may have used different methodologies for evaluating the denominator. However, because each firm evaluated the number of patient-years during the period for each anti-TNF agent, the difference in risk between agents that we observed cannot be explained by the different methodologies used for the different agents. Furthermore, in the sensitivity analysis, we used the different estimates from independent sources separately, which gave very consistent adjusted incidence rates and SIRs. In addition, our estimates for adalimumab and etanercept did not differ from data obtained from the RSI, which has claims data for outpatients receiving etanercept or adalimumab among 3 million RSI enrollees (i.e., 5% of the French population). The consistency of results of sensitivity analyses from these independent sources provides strong evidence for their validity. Furthermore, the difference in risk between agents that we observed (7–17-fold difference) is unlikely to be explained by a lack of precision in the denominator estimate. Although this national survey design has some limitations (especially the lack of precision in the denominator), it is probably the only way (or at least the most powerful way) to investigate difference in risk with use of anti-TNF agents.
The mechanism by which TNF antagonists reactivate latent TB is not fully understood. In animal models, TNF plays a central role in mediating mycobacterial infections, and sTNF (especially membrane-bound TNF) expressed by activated macrophages and T lymphocytes is essential in protecting against TB infection (18). In a murine model of chronic TB, neutralization of TNF by antibody but not the TNFR fusion molecule exacerbated chronic TB because of better penetration of antibodies into granulomas (19). In humans, higher avidity and better stability of membrane-bound TNF was demonstrated with anti-TNF mAb (20), which in some studies leads to more efficient apoptosis (21–23).
Even if in other studies infliximab and etanercept showed no clear difference in the ability to induce apoptosis in vivo (24), the difference in reverse signaling attributable to a difference in membrane-bound TNF targeting may have other consequences. For example, anti-TB–specific T cells show a clear difference in effector function. Two in vitro studies provided the same results: adjunctive infliximab or adalimumab treatment to specific human anti-TB T cells was much more efficient in decreasing proliferation and interferon-γ secretion by these T cells than was adjunctive etanercept treatment (20, 25). In addition, inhibition of interleukin-1 released by lipopolysaccharide-stimulated monocytes was greater with anti-TNF mAb treatment than with sTNFR treatment (26). Finally, anti-TNF mAb therapy alleviates the Treg cell impairment seen in RA by inducing apoptosis of membrane-bound TNF–positive inefficient Treg cells, thus allowing reexpansion of efficient Treg cells (27). Thus, our epidemiologic results can be supported by differences in the action of the 2 types of anti-TNF agents on specific effector T cells and Treg cells.
Two-thirds of our cases of TB occurred in patients with negative results of TB skin testing at screening. However, if the national recommendations for prophylactic treatment in France were strictly followed, 56.5% of the patients with a tuberculin skin test reaction diameter of <5 mm and 39.5% of those with a reaction diameter of <10 mm should have received an anti-TB prophylactic treatment. Concerning the others, either screening was not adequate or new TB infection developed instead of reactivation of latent TB. Our study suggests that taking into account whether patients were born in an endemic area could improve screening for latent TB. However, this information may have been less precisely reported in the control group. Other ways to improve screening for latent TB, such as use of specific in vitro blood tests assessing the presence of specific anti-TB T cells (QuantiFERON-TB or T-Spot.TB [Oxford Immunotec, Abingdon, UK]), should be evaluated.
Our study also confirmed that, as emphasized by other investigators (8), anti-TB chemoprophylactic treatment is efficient for preventing TB. None of our patients with TB had received a correct prophylactic antibiotic against TB. The efficacy of screening and anti-TB chemoprophylaxis may explain why the median time from the beginning of anti-TNF mAb therapy to TB onset is much higher (12 months and 10 months for the first and the last anti-TNF agent, respectively) in our study than in the first study describing TB in patients treated with infliximab (14 weeks) (9), when no screening for TB was recommended. However, we probably observed both features of TB infection in our study: reactivation of latent TB infection (corresponding to the high risk during the first year of anti-TNF treatment, as shown in Figure 1 and in the case–control results) and new infection occurring during followup, which will never be avoided with pretreatment screening.
In conclusion, our study provides clear evidence of a higher risk of TB with anti-TNF mAb therapy than with sTNFR therapy. These differences can be supported by differences in the action of the 2 types of agents on membrane-bound TNF, leading to a differential effect on effector T cells and Treg cells. These different mechanisms of action could also explain a better efficacy of mAb therapy in Crohn's disease, in other granulomatous diseases such as sarcoidosis (28), and in uveitis (29).
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Mariette had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Tubach, Salmon, Ravaud, Bréban, Chichemanian, Emilie, Lemann, Mariette.
Analysis and interpretation of data. Tubach, Salmon, Ravaud, Bréban, Chichemanian, Bretagne, Lemann, Mariette.
ROLE OF THE STUDY SPONSOR
Abbott, Schering-Plough, and Wyeth had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all of the data and had the final responsibility for the decision to submit the manuscript for publication. Publication of this article was not contingent on approval from Abbott, Schering-Plough, or Wyeth.
We are grateful to G. R. Auleley, J. Deligne, and C. Blum-Boisgard from the RSI for providing data to validate the denominator estimate of the incidence rate that we used, D. Che and D. Antoine from the Institut National de Veille Sanitaire for providing the French annual incidence rate of TB according to age and sex, and the AFSSAPS and the regional pharmacovigilance centers for their contribution to the exhaustiveness of the RATIO registry. We also thank C. Roy and G. Baron for the statistical analysis, and N. Nicolas, S. Makhlouf, and A. Djemoui for their help in collecting and preparing information validating the cases.